so we've got 8 SAS drives comin' and we want to run a zpool on them, but we also want to have swap and dump devices

[atax1a@infosec.exchange]

@nabijaczleweli @arrjay not true, according to the openzfs github? https://github.com/openzfs/zfs/pull/16906 / https://github.com/openzfs/zfs/issues/7734

[nabijaczleweli@101010.pl]

@atax1a hm. haven't encountered this even in what i'd expect to be pathological conditions. but maybe

[atax1a@infosec.exchange]

@nabijaczleweli yeah, it seems to be disrecommended. as we expect not to be running into swap very often, we'll just swap to two gmirrors and dump onto a dedicated dumpdev like it's 1998

[erincandescent@akko.erincandescent.net]

@atax1a @nabijaczleweli zvols are not very well maintained these days and perf is often worse than a file in a dataset, sadly, so I would disrecommend anything involving them

[atax1a@infosec.exchange]

we changed our mind on this, made 8x 128-gig swap partitions, mirrored those into 2 groups. no need for a dumpdev, anymore

[atax1a@infosec.exchange]

this fucking zpool scrubs at over 1 gigabyte per second, holy shit

[atax1a@infosec.exchange]

the old pool scrubbed in 16 hours, the new one is projected to take fucking 2

[astraleureka@social.treehouse.systems]

@atax1a SAS drives with big caches make things so much nicer, lol. back before fast SSDs were cheap I had 4x300G 10K SAS in my desktop in a RAID0 just for the sake of speed, even w/o array cache having 1GB/s sequential rw was :chefskiss:

[atax1a@infosec.exchange]

@astraleureka the caches on ours are 256 megs, comparable to the sata models, but the sata models don't give us extensive logs of every time their longitudinal error correction recovers a sector by rewrite

[astraleureka@social.treehouse.systems]

@atax1a SAS error reporting >>>>

[atax1a@infosec.exchange]

@astraleureka yeah all of these nerds on hobby sysadmin forums are like "buhh sas drives dont give smart data" and we're like, it's right there?? in the scsi log pages??

[astraleureka@social.treehouse.systems]

@atax1a oh my god even smartctl shows the scsi logs these days, how outdated is their brain history

also there's plenty of good drive debug tooling too, seagate open sourced seachest ages ago so low level poking at realloc history is trivial, even works with other vendors

[atax1a@infosec.exchange]

@astraleureka camcontrol on freebsd lets us dump the defects list and opcodes table and upgrade the drive firmware. openseachest is cute though